摘要
掌握巨量金融资源的金融控股公司一旦倒闭,不仅会带来巨额损失,甚至会引发金融危机。2008年金融危机中的"花旗事件"表明,监管机构针对金融控股公司构建有效的风险预警系统已刻不容缓。本文主要采用美国银行控股公司和台湾纯粹型金融控股公司作为研究样本,运用因子分析法对金融控股公司危机预警指标进行了降维,并采用BP神经网络方法实现了非线性指标自学习,构建了较具解释力的FA-BPNN金融控股公司危机预警模型。实证结果表明,FA-BPNN模型较多元判别分析具有更高的预测能力,对加强和改善金融控股公司的日常预警监管具有一定的参考意义。
Once a financial holding company,with huge financial resources,goes bankrupt,it would not only mean a huge amount of losses,but also result in financial crisis.The failure of Citigroup in 2008 demonstrated that the regulators had to build a risk early warning system immediately for financial holding companies.This paper uses the bank holding companies in America and the purely financial holding companies in Taiwan as research samples,and builds up the financial holding company risk early warning system combined with factor analysis and BP neural network,in which factor analysis is used for dimensionality reduction and BP neural network realizes its self-study functions on nonlinear mapping.Our findings demonstrate that the FA-BPNN model is much better in prediction ability than multiple discriminant models.This model could offer the regulators some references on the risk monitoring and prediction practice.
出处
《国际金融研究》
CSSCI
北大核心
2010年第11期37-46,共10页
Studies of International Finance
基金
国家自然科学基金(70673068)
同济大学人文社会科学重大课题培育计划项目的阶段性成果
关键词
因子分析
BP神经网络
金融控股公司
风险预警
Factor Analysis
BP Neutral Network
Financial Holding Company
Risk Early Warning
